Nvidia Transcript: Returning 50% of Free Cash Flow to Shareholders and the Rise of the Custom Vera CPU for Agentic AI
June 24, 2026 - Nvidia GTC Taipei 2026 Financial Analyst Q&A
Capital Return Program and the Agentic Computing Pattern
Presenter: With that, please join me in welcoming our founder and CEO, Jensen Huang.
Jensen Huang: I am coming. I am coming. Give me a second. Dajia hao, nice to see you. Did you see all the products we announced this week? Unbelievable. Okay, before we start, I have an announcement, a very exciting announcement. And it is not a secret. You can tell everybody. Tell all your friends. It is not an announcement about me. Everybody, okay, I will read it to you at earnings. We announced an $80 billion share repurchase program. You guys remember that? And a 25x increase in our dividend. You remember that? We also told you that we would review our dividend on a regular basis. So today I have a very exciting announcement that today we are announcing that we plan to return 50% or more of free cash flow to our shareholders this year, next year, and beyond. To infinity and beyond.
Jensen Huang: If you do some simple math, that is a lot of money. And we plan to increase our stock repurchases and our dividend over time. Okay, so that is our announcement. If you have any questions, we would be happy to take them. And so, this week, of course, yesterday, we announced some really exciting news. The big idea, the big idea, and I have been talking about it for two years, the big idea is that the computing pattern of AI is agentic. The computing pattern, how the agents are the modern applications. Agents are the modern applications. They can reason. They can use tools. They can access long-term memory. The long-term memory can be structured or unstructured. They could use tools. Those tools could be on your PC. It could be in the cloud. It could be design tools, software programming tools. It could be, of course, database retrieval, database processing. It could be used for designing chips using tools of all kinds. And so this computing pattern we call agents is going to run everywhere just as applications ran everywhere in the past. This computing pattern called agents will run in the cloud. It will run in the PC. It will run in workstations. It will run in the car, and will even run in humanoid robots.
Jensen Huang: And so I have been describing this, of course, piece by piece by piece over a long period of time. If you watch our GTC keynotes, you will see Jensen has been talking about this moment for two years, maybe longer, putting all the pieces together, okay? And this computing approach, this computing pattern is distributed and disaggregated. Disaggregated and distributed, which means that each part of the computing, the agent computing pattern, each part of that workload is running in different parts in the data center.
Designing Vera Rubin and Grace Blackwell for Agentic Workloads
Jensen Huang: Sometimes, that is the reason why we built Vera Rubin. Hopper was built for pre-training. Grace Blackwell introduced in addition inference. Grace Blackwell was designed for pre-training, post-training, as well as inference. And NVLink 72 made it possible for us to generate the lowest cost tokens in the world. Not by 20%, but by 20 times. And so we are now the lowest cost way to generate tokens. The goal is not to have a low-cost data center. The goal is to produce product with low cost. And so Nvidia's Grace Blackwell generates tokens at the lowest cost of anything in the world. We invented the NVLink scale-up switch as a result. Between NVLink, InfiniBand, and Ethernet designed for AI, we are now also the largest networking company in the world. So that is what happened with Grace Blackwell. Vera Rubin was designed for pre-training, post-training, inference, but also running agents, running this pattern, this computing pattern that I showed you over there yesterday. Vera Rubin was designed to run that whole thing, and that pattern, as I said earlier, is disaggregated and distributed. Disaggregated means different parts of that workload are running on different parts of Vera Rubin.
Jensen Huang: And some parts of Vera, the heavy part, the part that makes the money for the AI, is the token generation. That is the heavy part. That is why you want to maximize the number of GPUs you have, because what you sell is tokens. Does that make sense? You do not sell everything else. You are trying not to, you know, buy a lot of networking from us, but that is not your goal. You are going to buy a lot of storage, but that is not your goal. Your goal is to generate as many tokens as possible. That is the reason why GPUs will continue. You will start by saying, I have a 1 gigawatt data center, how many GPUs can I put in there? Because that maximizes my revenues.
Jensen Huang: Okay, now, but the middle is all thinking, and the thinking is very heavy compute. Thinking, reading the context, which is reading all the documents, reasoning about it, thinking, coming up with a plan, thinking, acting, which is generating commands for tools. That is thinking. Does that make sense? The tool, the tool feedback comes back. You look at it. You say, does that make sense? Is that the best outcome? Is the answer right? Now you are thinking again. Then you come back, ah, not quite right, I have to adjust my plan, use the tool again. Back and forth, back and forth, back and forth. And that is why you need the tool use to be very fast, because the AI is waiting. The AI is saying, here is the tool, I want you to use this tool, now I need the answer back as quickly as possible, which is the reason why we are working with the entire software industry to accelerate their applications. Adobe has accelerated their applications, they announced it yesterday, completely re-architecting Adobe Photoshop and Premiere for the first time in decades. We are accelerating Cadence, we are accelerating Synopsys, we are accelerating Ansys, we are accelerating Siemens, we are accelerating everything because the agents are impatient. Because if the agents are waiting, a 1 gigawatt data center of GPUs is waiting, you cannot wait. And so the CPUs cannot just be easy to rent, cheap to rent, they have to be fast to respond, which is the reason why single-threaded performance is important, super important. Single-threaded performance, not multi-threaded, not multi-core, but one CPU doing one job for one AI. Does it make sense? It has to be super fast. And because it is disaggregated, we also thought about how do we design the right CPU and where to put it? How do we design the right CPU and where to put it? And how do we integrate that into the Vera Rubin architecture?
Jensen Huang: We also created Vera Rubin so that it is the world's best data processor for memory systems. Remember, an AI has to have long-term memory. It has to have short-term memory. Memory is data. Moving data around all over the data center requires a lot of bandwidth. Which is the reason why Vera is also the first ConnectX-8. Vera has the most IO bandwidth. Vera also has the most CPU-to-CPU bandwidth because when you are doing data processing, the CPUs have to talk to each other. No chiplet tax, completely on one giant die. We could make it into four little dice, six little dice, but every time you cross the die, there is a chip tax. There is a chiplet tax. We do not want a chiplet tax. That is the reason why CPU-to-CPU bandwidth is three and a half times higher. The cross-sectional bandwidth inside our chip is absolutely the best in the world. The IO bandwidth is absolutely the best in the world, not by 15%, but by X factors.
Jensen Huang: Grace was the first CPU we decided to do this. Vera was the second generation. Our CPU core in Vera is completely custom because we wanted this CPU core, the CPU, to have the world's highest instructions per clock, IPC. Ten instructions at the same time per clock. We fetch 10, we decode 10, we execute 10 all the way through the pipeline. No CPU in the world can do that. As a result, you could see the performance. It is off the charts. Vera was not designed for humans. We are, you know, quite slow. Vera was designed for agents. Very impatient. Accelerated computing designed for agents. Very impatient, okay? And so this entire system was designed not just for pre-training, not just for inference, but to run agents, which is that schematic over there. That is the future. That is my attempt to take a very complicated system and turn it into a cartoon. But that pattern of computing is going to replicate everywhere. It is going to run in the cloud. That is what Vera Rubin is for.
Reinventing the PC Ecosystem with Windows and Sandboxed Agents
Jensen Huang: It is also going to run on a little desktop computer. And so several years ago, about three years ago, I talked to Satya Nadella and I said, in the future, AI wants to run on a device, too, because we want to have assistants running with us all the time. Right now, well, for you, all of your laptops are here. My laptop is in my room. If I want to talk to my laptop today, I have to wait until I get back to my room. In the future, if I need my laptop to do something, I just text it with WhatsApp. I say, you know, whatever, R2-D2, there is this thing with the PowerPoint slide, slide number 17. That image is scaled or titled wrong. It should not say CX9. It should say CX10. R2-D2 opens up PowerPoint, modifies it, puts it in PDF, and sends it to me. Can you imagine that? Easy.
Jensen Huang: That is a future where your AI is not just a laptop, not just a tool, but your laptop becomes an AI, becomes an assistant all day long. Now, you do not want to necessarily run everything in the cloud because if you can run it locally, it is free, just like laptops, just like phones, just like in your house. Why rent the television? You are going to use it every day. Why rent the washer-dryer? You are going to use it, well, hopefully once a week. Why rent your refrigerator? You are going to use it every day. Why rent your assistant computer? You are going to use it every day. Of course, we can also connect to a very powerful cloud, which is what Vera Rubin in the cloud is going to do. And so, you now have the benefit of assistants. You do not need to have just one assistant. You can have assistants of all kinds. And this assistant will orchestrate all of the other assistants. Of course, it would run on your desktop. It would run on your laptop. And so Satya and I, Microsoft and Nvidia, we decided to create a whole new line of computers. This entire line of computers, the first in the world that has tensor processing, that has parameter compression, that has an operating system that allows for a sandbox that is secure because you want to put your agents in a sandbox. Give it permission. Do not allow it to access everything in my laptop. You can access these files. You can use these tools. You can communicate with these people. Give your agent permission. You need an operating system to provide that sandboxing. And of course, we imagine the whole line of computers. In the past, we have workstations, we have desktops, we have laptops. So, of course, we need to have workstations, laptops, desktops, the entire line. And then we worked with Microsoft to make everything compatible.
Jensen Huang: In the last three years, we have been working around the clock. Everything is now compatible. And if it is not compatible, let me know, okay? And so everything that is important, every application that is important, we tested, performance-benchmarked it. It is an incredibly good experience. And then we invited the entire PC industry. The entire computer industry. Not one person left behind. Everybody is going to join us to reinvent the computer. So this is the first reinvention of the PC literally in 40 years. The behavior of your PC is going to change. It will do everything that it used to do better, and then it will do something else. It will also be your assistant. And so that was our announcement starting with the agentic computing pattern running in the cloud, in the data